Pine wilt disease, a highly contagious forest disease caused by the pine wood nematode and primarily transmitted via its insect vector, the pine sawyer beetle (Monochamus spp.), poses a significant threat to forest ecosystems. Accurate detection of infected trees is vital for effective prevention and control. This study pioneers the detection of pine wilt disease-infected trees in the China’s Qinba Mountain region, where the complex terrain and uneven forest distribution thinder feature extraction of diseased trees. To address data collection challenge, this paper proposes a novel feature recognition and detection method for pine wilt disease-infected trees based on an FLMP-YOLOv8 algorithm. The enhanced features include: first, integrating FasterBlock module into the backbone and neck networks of YOLOv8 to, boost the model’s feature extraction capability and reduce complexity, thereby achieving a balance between detection efficiency and accuracy. Second, a Large Separable Kernel Attention (LSKA) mechanism is incorporated into the Spatial Pyramid Pooling-Fast (SPPF) module of YOLOv8, improving the model’s ability to perceive fine details of diseased trees and reducing interference from other elements in the forest. Finally, the MPDIoU loss function is adopted for bounding box regression, enhancing the precision of localization. Experimental results on a self-constructed dataset demonstrate the improved model efficacy, achieving 92.0% precision, 80.8% recall, 87.0% mean Average Precision (mAP@0.5), and 81.79 FPS detection speed. Compared to the original YOLOv8 model, the improved algorithm shows increases of 2.2% in precision, 0.6% in recall, and 2.0% in mAP@0.5, with a detection speed improvement of 65.48 FPS. This study provides a more reliable and cost-effective method for detecting trees infected with pine wilt disease.

Pupil detection is a key technology in fields such as human–computer interaction, fatigue driving detection, and medical diagnosis. Existing pupil detection algorithms still face challenges in maintaining robustness under variable lighting conditions and occlusion scenarios. In this paper, we propose a novel pupil detection algorithm, ViMSA, based on the ViM model. This algorithm introduces weighted feature fusion, aiming to enable the model to adaptively learn the contribution of different feature patches to the pupil detection results; combines ViM with the MSA (multi-head self-attention) mechanism), aiming to integrate global features and improve the accuracy and robustness of pupil detection; and uses FFT (Fast Fourier Transform) to convert the time-domain vector outer product in MSA into a frequency–domain dot product, in order to reduce the computational complexity of the model and improve the detection efficiency of the model. ViMSA was trained and tested on nearly 135,000 pupil images from 30 different datasets, demonstrating exceptional generalization capability. The experimental results demonstrate that the proposed ViMSA achieves 99.6% detection accuracy at five pixels with an RMSE of 1.67 pixels and a processing speed exceeding 100 FPS, meeting real-time monitoring requirements for various applications including operation under variable and uneven lighting conditions, assistive technology (enabling communication with neuro-motor disorder patients through pupil recognition), computer gaming, and automotive industry applications (enhancing traffic safety by monitoring drivers’ cognitive states).

Let Y be a random variable whose moment generating function exists in a neighborhood of the origin. The aim of this paper is to study the probabilistic degenerate poly-Bell polynomials associated with the random variable Y, arising from the degenerate polyexponential functions, which are probabilistic extensions of degenerate versions of the poly-Bell polynomials. We derive several explicit expressions and some related identities for them. In addition, we consider the special cases that Y is the Bernoulli random variable with probability of success p or the gamma random variable with parameters 1,1.

Due to the different classification effects and accuracy of different classification algorithms in machine learning, it is inconvenient for scientific researchers to choose which classification algorithm to use. This paper uses the face data published by Cambridge University as an experiment. The experiment first reduces the dimensionality of the data through the principal component analysis (PCA) algorithm, extracts the main features of the data, and then respectively through linear logic classification, linear discrimination LDA, nearest neighbor algorithm KNN, support vector machine SVM and the integrated algorithm Adaboost are used for classification. By comparing the advantages and disadvantages of the classification performance and complexity of different algorithms, the final review reviews accuracy, recall, f1-score, and AUC as evaluation indicators.

Real-time pupil detection is a critical prerequisite for eye-tracking applications such as vestibular videonystagmography (VNG) and gaze-based interaction: deep learning methods offer high accuracy but incur substantial computational cost, while traditional thresholding methods are fast but lack robustness under complex illumination. This paper proposes the Confidence-Guided Fallback (CGF) framework, which employs four geometric factors—circularity, aspect ratio, area consistency, and convex-hull ratio—to evaluate the reliability of traditional detection results, and it invokes RITNet (a deep segmentation network finetuned on randomly cropped s-OpenEDS data) only when the confidence score falls below a threshold, thereby adaptively balancing accuracy and efficiency on a per-frame basis. RITNet is not trained on LPW; thus, LPW evaluation simultaneously constitutes cross-dataset validation. Hyperparameters (confidence weights and threshold values) are tuned on a validation set (15% of LPW subjects) and final performance is reported on an independent test set (18% of subjects), avoiding data leakage. On the validation set, the combined confidence factors achieve an AUROC of 0.961 for distinguishing high-quality frames, and the confidence distribution exhibits a bimodal structure with a valley near 0.7, providing a data-driven basis for threshold selection. On the independent test set (23,999 frames), CGF ($\tau =0.8$) detects 79.0% of frames; among detected frames, 84.0% achieve sub-5 px accuracy (median error 1.25 px) at 298 FPS on NVIDIA RTX 5090. At $\tau =0.7$, detection rate reaches 89.6% with 72.0% sub-5 px accuracy at 391 FPS. Failure-mode analysis reveals that approximately 64.3% of triggered deep-model frames represent “dilemma frames” where both traditional and deep methods fail, marking the practical ceiling of single-frame visual methods and pointing toward temporal modeling. The framework demonstrates robustness to confidence-weight perturbations (±20% weight changes yield < 1% performance variation) and exhibits strong discriminative power on high-confidence frames (93.4% accuracy within 5 px on the test set).